Google’s AI-Powered Image Rewriting: Revolutionizing Visual Content Enhancement

What’s Trending?

Google has introduced ImageFX, an AI-powered tool for picture generation, and deepfakes of Taylor Swift aren’t going away anytime soon.

Use ImageFX’s prompt-based user interface to make and edit images; it’s built on Imagen 2, a GenAI picture model created by Google’s DeepMind team. That’s the same as programs like Microsoft Designer, Meta’s Imagine with Meta AI, Midjourney, and OpenAI’s DALL-E 3. However, what really sets ImageFX apart are its “expressive chips”—essentially, a collection of keyword suggestions that enable users to play around with the “adjacent dimensions” of their works and ideas.

Read: Top 10 News of Google in 2023

According to Google’s blog post, ImageFX is “designed for experimentation and creativity.” It allows users to make images using a basic text prompt and then simply change them using expressive chips, which is a new way of prompting.

What About the Other Possibilities?

Google asserts that it has implemented measures to prevent ImageFX from being misused, such as implementing “technical safeguards” to restrict “problematic outputs” such as sexually explicit, violent, offensive, and insulting material. Although Google was vague about it in its promotional materials, ImageFX also includes a prompt-level filter for “named people,” who are likely celebrities.Also today, Google said that more of its products and services, including as its next-gen AI search experience and family of managed AI services Vertex AI, will begin to use Imagen 2 this week.

Read: 10 AI In Manufacturing Trends To Look Out For In 2024

Imagen 2, a component of Google’s GenAI suite of products for productivity, has found its way into SGE (Search Generative Experience). It also supports text-to-image capabilities in Google Ads and Duet AI in Workspace. Imagen 2 is now used by SGE to generate images. Last October, SGE started surfacing image generation tools for consumers in Google Image Search. In the SGE conversational experience, users can command SGE to provide four options by entering a prompt describing the type of image they desire.